Written by: Steven Silverman (@silver_stats)
The 11th annual Sloan Sports Analytics Conference took place in Boston from March 3-4, featuring as usual a wide variety of panelists, research papers, competitions, and conversations. Below is a brief recap highlighting some of the best moments of the weekend.
Hackathon: On Thursday, March 2nd, the day before the conference proper, 47 entrants participated in the third iteration of the Sloan hackathon, co-sponsored by ESPN and Ticketmaster. The theme was “measuring the immeasurable,” with SportVU player tracking data from five NBA games given to the participants to delve into. Some of the more interesting projects included measuring how much of “good soldiers” players were by looking at unselfishness, how hard individual players crash the boards, and whether or not getting scored upon has an effect on defensive effort on a subsequent possession. The University of Pittsburgh’s Kostas Pelechrinis, who has been partnering with TSAC during the last year, was a finalist in the open division for his work on calculating team chemistry by using concepts from network theory, like fractal dimension and algebraic connectivity.
Research Paper Competition: The top eight research paper submissions, from among the many entries, were given the chance to present their findings to the conference. The winning paper, “From Sports to Science: Using Basketball Analytics to Broaden the Appeal of Math and Science Among Youth,” examined the results of several clinics the authors had run in which youth basketball players were given the chance to track their own stats and improve their games, then surveyed (both before and after) about their attitudes toward STEM education and their willingness to consider STEM as a career. Other finalist papers included “Possession Sketches: Mapping NBA Strategies,” which took advantage of natural language processing techniques to interpret each possession as a “document” and classify it accordingly, and “Data-Driven Ghosting Using Deep Imitation Learning,” which sought to predict how soccer possessions would have been defended by different teams to inform future strategy.
Win the (Game) Day: Featuring Carnegie Mellon professor and TSAC advisor Sam Ventura, this panel discussed how analytics can be used during a game to make decisions and adjustments. All of the panelists, among them Denver Broncos director of analytics Mitch Tanney and Ringer author Ben Lindbergh, stressed that mutual respect and trust are essential for the coaching staff and players to have a good relationship with their analysts. They also discussed the limitations of in-game adjustments for decisions that only appear infrequently, as well as the job security incentives for coaches to remain conservative.
Moneymind: As described by moderator Cade Massey, this “panel of illustrious general managers—and Sam [Hinkie]” talked about the dangers of cognitive bias in sports decision-making. A’s GM Billy Beane, Dodgers GM Farhan Zaidi, and Rockets GM and SSAC organizer Daryl Morey traded banter about the importance of applying the proper weighting to the various inputs and decision-makers influencing you, as well as specific strategies like mentally reversing a trade to evaluate it and minimizing how big a role regret plays in a decision.
Ball Don’t Lie: A basketball analytics panel consisting of Sue Bird, Luis Scola, Celtics AGM Mike Zarren, TruMedia VP Dean Oliver, former Bulls and Clippers coach Vinny Del Negro, and moderator Zach Lowe, provided a wide range of viewpoints. Bird and Scola both discussed the importance of tailoring information to specific players and not overwhelming them with facts, and how for certain rookies, one or two focal areas could be too much, while a veteran could handle a dozen or so. Oliver stressed that tests themselves need to be tested sometimes, and that teams should be careful draft prospects aren’t just succeeding at the test, but that their skills will carry over into games.
Caught in a Pickle: The yearly baseball analytics panel came on the heels of a presentation by Mike Petriello and Greg Cain of MLBAM unveiling new Statcast metrics for 2017, including “catch probability” and “hit probability” on nearly every batted ball. Panelists Petriello, Dave Cameron, Harry Pavlidis, and Patrick Young and moderator Brian Kenny talked about how the online baseball research community has directly impacted individual players, especially with the emergence of Statcast, and how many ex-players, like John Baker and Dan Haren, have turned to analysis as a second career. They also discussed how the baseline of information understood by the average fan has risen over the last decade or so, how the average age of managers is falling rapidly, and the importance of diversity in hiring (and, importantly, how to achieve such diversity when the applicant pool is skewed by external factors).
Other Events: Other interesting panels included Nate Silver and Adam Silver speaking one-on-one; Nate Silver and Mark Cuban speaking one-on-one; Shane Battier, Marcus Lattimore, and others discussing the life of a college student-athlete in the digital age; John Urschel on a football analytics panel focusing on stopping punting; a sports science panel discussing wearables, athlete monitoring, and the ethical questions that arise from doing so; a soccer analytics panel with excellent insight on how getting game-style data during training can greatly increase confidence in decision-making; and a panel on “Sustaining Greatness” that featured Sue Bird and chess grandmaster Susan Polgar. The conference also featured more content on eSports than in years past, including Smash 4 tournaments (1v1 pro and 2v2 pro-am), as well as a drone flying demonstration.
All in all, it was a very successful conference, and we look forward to having more and more TSAC members attend in years to come!